"""
Image registration based on the radon transform.
Author: Fabian A. Preiss
"""
import numpy as np
from imgreg.models.radon.params import (
RefImageParam,
ModImageParam,
AngularPrecisionParam,
ExponentialFilterSignalNoiseParam,
ThetaParam,
UpsamplingParam,
RefRotationParam,
ModRotationParam,
RotationCandidateParam,
AngleSelectParam,
RecoveredRotationParam,
RecoveredRotParam,
RecoveredTranslationParam,
RecoveredRotTrParam,
)
from imgreg.models.radon.enums import RadonParams
from imgreg.util.solver import Solver
from typing import Optional
[docs]class RadonSolver(Solver):
r"""
Implements an image registration model based on the radon transform.
The model tries to reconstruct the difference of rotation and translation between two images.
The model is limited such, that due to the symmetry of the method it only considers
angles between 0 and 180 degrees. It will fail, if the rotation angle is larger.
Parameters
----------
ref_img : numpy.ndarray
The original input image (one color channel only).
mod_img : numpy.ndarray
The modified input image (one color channel only).
Notes
-----
The model implements the following dependency graph to construct it's `Parameter`\ s.
.. graphviz:: ../_assets/dependencies_radon.gv
The `Parameter`\ s are documented in :ref:`imgreg.models.radon.params`.
Examples
--------
We can visualize the internal `ImageParameter`\ s of the model as follows:
.. plot:: tutorial/pyplots/radonsolver.py
:include-source:
If we simply want to create a model and access the recovered values we first setup a model:
>>> import numpy as np
>>> import imgreg.data as data
>>> from imgreg.models.radon import RadonSolver
>>> ref_img = np.array(data.ref_img())
>>> mod_img = np.array(data.mod_img())
>>> ras = RadonSolver(ref_img, mod_img)
Now the parameters of the model can now be accessed as follows:
>>> ras.RECOVERED_ROTATION.value
array([-11.35802469, 0.28284271])
>>> ras.RECOVERED_TRANSLATION.value
array([-11.80822224, 25.77936941, 0.30466765])
"""
def __init__(
self, ref_img: Optional[np.ndarray] = None, mod_img: Optional[np.ndarray] = None
):
super().__init__()
p_ref_img = RefImageParam(ref_img)
p_mod_img = ModImageParam(mod_img)
p_angular_precision = AngularPrecisionParam(0.2)
p_exponential_filter_signal_noise = ExponentialFilterSignalNoiseParam(100)
p_theta = ThetaParam(np.arange(0, 180, 1))
p_upsampling = UpsamplingParam(10)
p_ref_rotation = RefRotationParam(
{p_ref_img, p_theta, p_exponential_filter_signal_noise, p_angular_precision}
)
p_mod_rotation = ModRotationParam(
{p_mod_img, p_theta, p_exponential_filter_signal_noise, p_angular_precision}
)
p_rotation_candidate = RotationCandidateParam({p_ref_rotation, p_mod_rotation})
p_angle_select = AngleSelectParam(
{p_mod_img, p_ref_img, p_rotation_candidate, p_upsampling}
)
p_recovered_rotaton = RecoveredRotationParam(
{p_rotation_candidate, p_angle_select}
)
p_recovered_rot = RecoveredRotParam({p_angle_select})
p_recovered_translation = RecoveredTranslationParam({p_angle_select})
p_recovered_rot_tr = RecoveredRotTrParam({p_angle_select})
self._register_params(
{
p_ref_img,
p_mod_img,
p_angular_precision,
p_exponential_filter_signal_noise,
p_theta,
p_upsampling,
p_ref_rotation,
p_mod_rotation,
p_rotation_candidate,
p_angle_select,
p_recovered_rotaton,
p_recovered_rot,
p_recovered_translation,
p_recovered_rot_tr,
}
)
self._generate_dep_graphs()
for key in RadonParams:
setattr(self, key.name, self[key])
#